Communications Department

Analysis of the competition between operators

in a femtocell scenario
(Document TELPOL73)
Authors: Luis Guijarro Outreach: Public
Vicent Pla, Jose R. Vidal Date: October 24, 2012
Jorge Martinez Bauset Version: d
1 Objective
This paper analyses the eect of the entry of a femtocell operator in a mobile commu-
nications market where an macrocell operator exists.
The analysis is conducted using a game theory based model.
2 Problem statement
The problem is part of the implementation of the work programme nanced by the
Ministerio de Economa y Competitividad within the research project COHWAN. One
of the project objectives is:
To analyse through game theoretical modelling the economic aspects of the spec-
trum sharing. The analysis will depart from the framework provided by the
cognitive radio networks ...
The problem stated was to answer the following question: which benet do mobile
communication users get from the entry of femtocell operators into the market?
The basic scenario comprises the following elements:
Macrocell Operator (MO), which provide service to mobile communication users
and which are entitled to use the spectrum for the provision
Femtocell Operator (FO), which provide service to mobile communication users
and which exploit an amount of spectrum owned by the Macrocell Operator.
The latter lease spectrum to the former in order to improve the eciency of the
spectrum use and to increase their revenue.
Users, which are the mobile communication users and which receive the service
provided either by an MO or by an FO.
As regards the femtocell deployment, the following assumptions are made
1
:
The femto base stations (femto-BSs) are deployed by the FO and no capital nor
operational expenses are levied onto the users. The MO is not involved in the
femto-BS deployment, either.
The femto-BSs operate in open-access mode, as this mode allows the FO to oer
service to users
An orthogonal spectrum assignment is performed through MO-FO coordination,
so that cross-tier interference is completely eliminated.
The concrete setting that will be analysed is shown in Fig. 1, where:
One MO and one FO compete for the provision of service to the users. The MO
and the FO are also referred to as operator 1 and operator 2, respectively.
No switching costs are incurred when users make the subscription decision
1
For an introductory description of the deployment alternatives in femtocell networks, the reader
may refer to Lin, Zhang, Chen, and Zhang (2011)
2
MO
2
1
FO
FO
Figure 1: Scenario
The FO have the same service coverage as the MO, and each user can access at
every point in space to both MO and FO services. This assumption is also made
by Duan and Huang (2011) and will be relaxed in future works.
The MO will lease an amount b of spectrum to the FO, keeping for itself the rest,
up to W. The FO pays a price p m.u.
2
per unit of spectrum. Finally, a user would pay
p
1
m.u. if she subscribed to MO service, or p
2
m.u. if she subscribed to FO service.
All three prices p, p
1
and p
2
are referred to the same time period.
3 Literature review
There have been recent works on economic modelling of femtocell service provision:
Yi, Zhang, Zhang, and Jiang (2012) covers a similar setting as in our current
work, since it analyses the spectrum leasing between a macrocell service provider
and a femtocell service provider. It is concerned not only with the amount of
spectrum which is leased, but also with the amount of leased spectrum that
the femtocell service provider is willing to share with the users of the macrocell
service provider. However, Yi, Zhang, Zhang, and Jiang (2012) builds a non-
standard model according to microeconomics, since they do not incorporate the
2
m.u.= monetary units
3
users demand in the model; instead, they insert a proxy for the users utility in
the service providers utility.
Duan, Shou, and Huang (2011) models also a scenario where one macrocell op-
erator leases spectrum to one femtocell operator, and it derives the conditions
under which the former has incentives to lease. It assumes, as in our paper, that
the femtocell operator does not incur costs and that it provides the same coverage
as the macrocell operator, but it also examines the cases where each assumption
does not hold. However, Duan, Shou, and Huang (2011) assumes that the leas-
ing (wholesale) price and the service (retail) price are the same and it bases the
assumption on the need to avoid arbitrage. We strongly argue against the above
assumption and we therefore model the leasing price and the subscription price
in separate stages.
4 Model
We assume that the operators compete `a la Bertrand, that is, playing a one-shot si-
multaneous game where MO and FO strategies are p
1
and p
2
, respectively.
Both the price p and the leased spectrum amount b are partially the result of a
bargaining process between the MO and the FO.
Finally, each user will subscribe to the service providing the highest utility. The
subscription period matches the time period of the bargain price p and the oered
prices p
1
and p
2
. Assuming that the number of users (n) is high enough, the individual
subscription decision of each user will not aect the utility perceived by the rest, and
a Wardrop equilibrium will result.
Thus, we model the strategic interaction between the two operators and the n users
as a three-stage multi leader-follower game, which comprises three phases:
First phase: the bargain between the MO and the FO takes place, which may result
in an outcome which gives a set of price p and amount b of the leased spectrum.
Second phase: the MO oers a subscription price p
1
and the FO oers p
2
.
Third phase: each user chooses whether to subscribe to the service or not; if she
chooses to subscribe, she simultaneously chooses which operator to subscribe to.
The number of users that subscribe with operator i is denoted by n
i
.
A standard way to analyse this sort of games is by means of backward induction.
4.1 Operators prot
The operators prots can be expressed as:
Macrocell operator

1
= n
1
p
1
+ p b C
1
(1)
Femtocell operator

2
= n
2
p
2
p b C
2
(2)
where C
i
is the costs beared by operator i.
4
4.2 Subscription
In this phase, a pair of values b and p has been agreed and the prices p
1
and p
2
have
been announced.
The utility that the users receive from each operator depends on two factors:
Quality of service Each operator exploits, during each subscription period, an amount
of spectrum which is agreed at the end of the rst phase: W
2
= b for the FO and
W
1
= Wb for the MO. Dierent users experience dierent channel conditions to
the macrocell base stations due to dierent locations, and thus achieve dierent
data rates when using the same amount of bandwidth. We model this fact by a
macrocell spectrum eciency .
When using the femtocell service, since femtocell base stations are deployed in-
doors and are very close to the users cell phones, we assume that all users using
the femtocell service have equal good channel conditions and achieve the same
maximum spectrum eciency.
Bearing in mind the above issues, the product of the spectrum times the spectral
eciency for each operator, divided by the number of users which subscribe to it
will give the transfer rate that is oered to each user (Niyato and Hossain 2008).
We propose to use this transfer rate as the main quality factor which contributes
to the user utility.
Price: the higher the subscription price, the lower the user utility.
Based on the above discussion, we propose a quasi-linear expression for the user
utility. Specically, for operator is subscriber, it is
U
i
= log

1 +

i
W
i
n
i

p
i
. (3)
where
2
= 1, for the FO, and
1
<
2
for the MO.
Each user will subscribe to the operator providing the service with the higher utility.
Assuming that the number of users is high enough, the individual subscription decision
of each individual user will not aect the utility received by the rest. Then, the equi-
librium notion is the Wardrop equilibrium, where each user makes her decision so that
she is indierent to choosing one alternative among all the available alternatives, and
therefore no user has an incentive to change her decision.
Each user has three subscription alternatives:
1. to subscribe to the MO,
2. to subscribe to the FO or
3. not to subscribe to either.
We assume that the users who do not subscribe have a utility equal to 0. Applying
the Wardrop equilibrium concept, we may state the following:
if some users decide not to subscribe to either the MO or the FO, then U
1
= U
2
=
0, i.e., the users are indierent between subscribing or not.
5
alternatively, if every user subscribes to either the MO or the FO, then U
i
0
if U
1
> U
2
, then no user will subscribe to the FO, and all users will subscribe
to the MO
if U
1
= U
2
, then users will distribute between the MO and the FO
if U
1
< U
2
, then no user will subscribe to the MO, and all users will subscribe
to the FO
We address the rst scenario, and show that actually it cannot occur in the equi-
librium. Let us assume that U
1
= U
2
= 0 and some users do not subscribe to either
operator. Then, from (3) it follows that at the user equilibrium, the number of sub-
scribers n
i
will be given by
n
i
=

i
W
i
e
p
i
1
. (4)
It can be observed that the number of subscribers each operator obtains solely
depends on its subscription price and not on that of the competitor. This fact can
be explained as follows. If one operator lowered the price p
i
, it would attract users
who would not have subscribed to the service otherwise, but it would not attract any
subscriber from the competitor.
Anticipating the above user equilibrium, each operator will choose the value of p
i
so that its prots are maximized. From (1)(2) and (4) we have that

i
p
i
=
i
W
i
(1 p
i
)e
p
i
1
(e
p
i
1)
2
< 0 , if p
i
> 0 ,
which indicates that by lowering the price, the operator gets more customers and in-
creases the prots. Hence, the operator would prefer to lower the price as long as there
are non-subscribing users, which eventually leads to the rst scenario boiling down to
the second scenario, where all users subscribe to the service.
The second scenario will be assumed in the next sections. As regards the three
alternatives enumerated above, we will restrict in next section to the case where U
1
=
U
2
. The case U
1
> U
2
where no user subscribes to the FO cannot be an equilibrium,
apart from the trivial case b = 0. From (3), and assuming nite prices, the utility U
2
that a user will potentially obtain from switching to the FO will be arbitrarily large.
But this contradicts the fact that U
2
is bounded by U
1
. Analogously, the case U
2
> U
1
where no user subscribes to the MO cannot be an equilibrium, apart from the trivial
case b = W.
4.3 Competition
In this phase, a pair of values b and p has been agreed, and each operator chooses
its strategy, which is the subscription price p
i
, aiming to maximise its prots. Each
operator does not know the strategy chosen by its competitor, but common knowledge
of the strategies available to them and the corresponding prots is assumed.
Here it is assumed that we are under scenario 2 as dened above, that is, all users
subscribe with one of the operators (n
1
+ n
2
= n) and both operators have customers
(U
1
= U
2
). Let us denote by = n
1
/n the fraction of customers that subscribe to
6
operator 1. We can then express the number of customers with each operator as a
function of
n
1
= n (5)
n
2
= (1 )n. (6)
To simplify notation we introduce
r
i
=

i
W
i
n
, i = 1, 2 . (7)
Using the equilibrium equation (U
1
= U
2
)
log

1 +
r
1

p
1
= log

1 +
r
2
(1 )

p
2
(8)
and (5)(6), n
1
and n
2
can be expressed as functions of p
1
and p
2
, so that:

1
=
1
(p
1
, p
2
)
2
=
2
(p
1
, p
2
). (9)
The equilibrium strategies p

1
and p

2
will be given by the Nash equilibrium condi-
tions:

1
(p

1
, p

2
)
1
(p
1
, p

2
), p
1
, (10)

2
(p

1
, p

2
)
2
(p

1
, p
2
), p
2
. (11)
Under the assumption that the partial derivatives of
1
and of
2
exist, the Nash
equilibrium will be among the solutions of the equation system:
0 =

1
(p

1
, p

2
)
p
1
= n

+ p

p
1

, (12)
0 =

2
(p

1
, p

2
)
p
2
= n

(1 ) p

p
2

, (13)
which yields
p
1
=

p
1

1
(14)
p
2
= (1 )

p
2

1
. (15)
Now, by taking derivatives with respect to p
1
and p
2
in (8) we obtain

p
2
=

p
1
=

2
/r
1
+
+
1
(1 )
2
/r
2
+ 1

1
. (16)
Combining (16) with (14)-(15) gives
p
1

=
p
2
(1 )
=
1

2
/r
1
+
+
1
(1 )
2
/r
2
+ 1
. (17)
7
From here p
1
and p
1
can be expressed as functions of , which substituted into (8) give
log

1 +
r
1

2
/r
1
+
+
1
(1 )
2
/r
2
+ 1

=
log

1 +
r
2
(1 )

(1 )

2
/r
1
+
+
1
(1 )
2
/r
2
+ 1

. (18)
Introducing
f(x, a, b) = log

1 +
a
x

1
x
2
/a + x
+
1
(1 x)
2
/b + 1 x

(19)
we can rewrite (18) as
f(, r
1
, r
2
) = f(1 , r
2
, r
1
). (20)
It is not dicult to check that as varies in the interval (0, 1) f(, r
1
, r
2
) monotonically
decreases from to and f(1 , r
2
, r
1
) monotonically increases from to .
Therefore, there exists a unique value

(0, 1) which satises (18).

The analysis of equilibrium in the competition setting can be formulated more
generally as a pair of optimization problems

max
,p
1
f
1
(, p
1
) = p
1
subject to g
1
(, p
1
) = U
1
U
2
= 0
h
1
(, p
1
) = U
1
0
and

max
,p
2
f
2
(, p
2
) = (1 )p
2
subject to g
2
(, p
2
) = U
1
U
2
= 0
h
2
(, p
2
) = U
2
0
Note that we have introduced as an auxiliary optimization variable. In the second
phase of the game the strategy of each operator consists of only its subscription price.
The constraint U
1
= U
2
denes as a (decreasing) function of p
1
for a given value of
p
2
, and as a function of p
2
for a given value of p
1
.
The KarushKuhnTucker (KKT) conditions for those two problems are
f
1
+
1
g
1
+
1
h
1
= 0 (21)
g
1
= 0 (22)
h
1
0 (23)

1
h
1
= 0 (24)

1
0 (25)
and
f
2
+
2
g
2
+
2
h
2
= 0 (26)
g
2
= 0 (27)
h
2
0 (28)

2
h
2
= 0 (29)

2
0 (30)
In this section we have so far implicitly assumed that U
1
= U
2
> 0, so that the
inequality constraints are not active and the optimization problems are the ones we
8
already solved and whose solution is given by (18) and (14)-(15). That solution, how-
ever, yields U
1
= U
2
< 0 for some settings and in that cases the inequality constraint
cannot bet let out. Now we examine those cases with active inequality constraints, i.e.,
h
1
= 0 or h
2
= 0.
If h
1
= 0, from (23) it follows that
p
1
= log

1 +
r
1

, (31)
which plugged into (22) yields
p
2
= log

1 +
r
2
1

. (32)
Finally, from (21) we obtain the following condition
f(, r
1
, r
2
) =

1
(1 )
2
/r
2
+ 1
0, (33)
where the function f is dened as in (19).
Likewise, if h
2
= 0 from (26)(30) we obtain again Eqs. (31)-(32) and the condition
f(1 , r
2
, r
1
) =

2

2
/r
1
+
0. (34)
Let
1
and
2
be, respectively, the only solutions in (0, 1) of f(, r
1
, r
2
) = 0 and
f(1 , r
2
, r
1
) = 0. As noted above, f(, r
1
, r
2
) is monotonically decreasing from
to and f(1 , r
2
, r
1
) monotonically increasing from to so
1
and
2
are
well dened. Moreover, condition (33) holds if, and only if,
1
; and (34) holds if,
and only if,
2
.
Depending on the relative position of
1
and
2
we have one of the following three
situations:
If
1
>
2
it is not possible to meet conditions (33) and (34) simultaneously.
Hence, there does not exist an equilibrium such that U
1
= U
2
= 0. Furthermore,
the solution to (18)

(
2
,
1
) yields U
1
= U
2
> 0.
If
1
<
2
any (
1
,
2
) satises (33) and (34), which substituted into (31)
and (32) yields a pair of prices that is an equilibrium point. Therefore, there
exists an innite and non-denumerable set of equilibrium points.
If
1
=
2
, then

=
1
=
2
is both the only solution to (18) and the only
value that satises (33) and (34) simultaneously.
4.4 Operators bargaining
As stated at the beginning of the section, the price p and the amount of spectrum b
are subject to a bargaining between the MO and the FO, which is conducted before
the subscription prices are advertised by the operators and the subscription decision is
made by the users.
Following Mukherjee and Pennings (2006), we model the bargaining as a non-
cooperative game where the incumbent operatorthe MOhas full bargaining power
9
and therefore oers a take-it-or-leave-it oer to the entrant operatorthe FO. The
oer consists of a pair of values p and b which the FO may accept or refuse. If the FO
accepts the oer, then the competition and subscription phase do take place; else, the
FO does not enter into the market, and the MO remains as the monopolist operator.
The bargaining is then proposed to be modeled as a dynamic game in a extensive
form:
1. The MO operator oers a pair of values p and b
2. The FO operator may accept or refuse the oer.
(a) If the FO accepts it, the operators will obtain prots which depends on the
competition and the subscription phase, as modeled above and denoted as

1
and
2
.
(b) If the FO refuses it, the FO does not enter into the market and therefore
obtains zero prots, and the MO remains as the monopolist operator and
obtains prots denoted as
m
.
Following backward induction, depending on the values of (b, p) which characterise
the MO oer, the bargaining outcome will be one or the other:
1. In order that the MO makes an oer, it must indeed prefer the competition
outcome rather than the monopolist outcome, that is,
1

m
.
2. If the oer made by the MO complies with

2
0, (35)
the FO will accept the oer. Otherwise,
2
< 0, and the FO will refuse it.
To sum up, the FO operator will enter into the market and competition will be
observed in the service provision to the users if the oer (b, p) made by the MO operator
is such that:

1

m
and
2
0, (36)
In other words, that both the MO and the FO are better o than it would be the case
if the former remains as the monopolist and the latter refrains from entering into the
market.
4.5 Monopoly prots
In the previous section, when deciding which oer to make to the FO, the MO compared
prots in the competitive environment with prots in the monopolistic case. Thus, we
need to compute
m
.
In order to compute the prots
m
, the expressions for the prots (1) and the user
utility (3) will be computed taking b = 0 and substituting p
m
for p
1
:
Prots are now given by

m
= n
1
p
m
C
1
. (37)
10
User utility is given by
U
1
= log

1
W
n
1

p
m
. (38)
To compute
m
, the problem should be stated as an optimal decision problem, such
that the optimal price p

m
should fulll:

m
(p

m
)
m
(p
m
) , p
m
.
From the perspective of the MO the situation here is, in a sense, similar to the
rst scenario we ruled out in Section 4.2, where there would have been no competition
between operators. There we concluded that the best interest for the operator was to
get as many customers as possible by setting the price appropriately. Applying the
same reasoning here, we obtain that the optimal price is
p

m
= log

1 +

1
W
n

, (39)
and then we obtain the value for the monopoly prots.

m
(p

m
) = nlog

1 +

1
W
n

C
1
. (40)
We assume that the monopoly prots are positive. Note that
m
(p

m
) is increasing
with n and
lim
n

m
(p

m
) =
1
W C
1
. (41)
Consequently,
1
W > C
1
is a necessary condition for the operator being able to make
positive prots (
m
(p

m
) > 0). Moreover, even if that condition holds, a minimum
number of users n is still required so that the operator makes some prots.
4.6 Bargaining outcome
From the conditions for the competitive equilibrium (see equation (36)) and using the
expression for the monopoly prots in (40), we will proceed to dene the feasibility
region of values (b, p) which allows for the competitive equilibrium to result.
From the condition
2
0, we derive:
p U(b) =
n
b
((1 )p
2
C
2
/n) , (42)
and from
1

m
:
p L(b) =
n
b

log(1 +

1
W
n
) p
1

. (43)
Note that in the above expressions for L(b) and U(b) the values of , p
1
and p
2
do also
depend on b.
Therefore, for an amount of leased spectrum b competition could occur if and only if
L(b) U(b), (44)
11
which using (42) and (43) can be written as
p
1
+ (1 )p
2
log(1 +

1
W
n
) C
2
/n. (45)
Furthermore, a leased spectrum of b at a price p will lead to competition, i.e. the
point (b, p) will be in the feasibility region, if and only if,
max

0, L(b)

p U(b). (46)
The nal bargaining outcome will depend on the specic assumptions made over
the bargaining process. If the incumbent has full bargaining power, as stated above
in 4.4, then the value p will be such that the equality holds in (42), i.e.
p = U(b) (47)
Given that
1
is monotonically increasing on the value p (see equation (1)), this would
provide the incumbent with maximum prots.
By noting that

p=L(b)
=
m
and (48)

p=U(b)
= 0 , (49)
we can rewrite (1) and (2) as

1
=
m
+

p L(b)

b , (50)

2
= U(b) p , (51)
from what it follows that

1
+
2
=
m
+ , (52)
where =

U(b) L(b)

b 0 is the amount by which the total prot is incremented

if the FO enters the market. Note that does not depend on p. Note also that the
problem of deciding how this extra prot is shared between the MO and the FO is
equivalent to setting the price p at which the MO sells bandwidth to the FO.
For this problem multiple solution concepts can be borrowed from the cooperative
game theory. For instance, the situation when the FO enters the market by leasing
some bandwidth from the MO may be regarded as a coalition between the two. The
total payo of the coalition will be
m
+ . Therefore, by agreeing to share both
players will be better o than if each of them plays by itself, where the prot of the MO
will be
m
and that of the FO 0. Note that in this context the MO and the FO forming
a coalition means that both agree on a price p but from that point on they compete in
a non-cooperative fashion in the retailer game by setting their subscription fees p
1
and p
2
independently from each other. In this context, the Shapley value provide a
fair allocation of the payo obtained by the coalition. In our case, the Shapley value
allocation yields

1
=
1
2

m
+
1
2
(
m
+ 0) =
m
+

2
, (53)

2
=
1
2
0 +
1
2
(
m
+
m
) =

2
, (54)
12
which correspond to
p =
U(b) + L(b)
2
. (55)
Alternatively, the problem of agreeing a value for p can be casted into a two person
bargaining problem in which the disagreement point is (
m
, 0) and the players strate-
gies are their oers about p. In this setting, both the Nash bargaining solution and the
Kalai-Smorodinsky bargaining solution yield the same results as the provided by the
Shapley value (see (53) and (54)).
Note, however, that the value b is not determined by the bargaining, but only
constrained by (44) .
5 Solution
Based on the model developed in the previous section, we propose the following solving
procedure:
1. The following parameters are xed: n, W, C
1
, C
2
,
1
, and
2
.
2. We propose a value b, such that 0 < b W, and assume that (45) holds.
3. We compute the solutions to (33) and to (34). Only if both solutions are equal,
there exists a unique

solving (18), and we proceed to compute the equilibrium

values.
4. We obtain the equilibrium strategies for each operator p
1
and p
2
, using (17).
5. Condition (45) is tested.
(a) If it is false, the outcome is that MO remains as monopolist and FO does
not enter:
i. For the MO operator,
A.

= 1
B. p

1
= p

m
, as given by (39)
C.
1
=
m
, as given by (40)
ii. For the FO operator, p

2
= 0, and
2
= 0
(b) If it is true, we set p

i
= p
i
and compute U
i
. Furthermore, a value of p is
chosen according to the assumptions made in section 4.6. Prots for MO and
for FO,
1
and
2
, are computed using expressions (1) and (2), respectively.
In order to evaluate the dierent competitive equilibria, we propose to use the
following indicators:
1. Operators prots
1
and
2
,
2. User utilities, U
1
= U
2
3
,
3. User welfare, UW, computed as the aggregated user utility over the total number
of subscribers:
UW = n U
1
= n U
2
(56)
3
This equality holds for the competitive equilibrium
13
Parameter Value
n 10000 users
W 95 kHz
C
1
20 m.u.
C
2
10 m.u.

1
0.5 bit/s/Hz

2
1 bit/s/Hz
Tabla 1: Parameters
4. Social welfare, SW, computed as the sum of the user welfare and the producer
welfare:
SW = UW +
1
+
2
(57)
6 Results
In this section, the numerical results obtained when following the solving procedure
outlined in the previous section are presented. First, the parameter values are listed.
Second, the equilibrium for each parameter conguration is represented in graphs. And
third, the results are analysed.
6.1 Parameters
The values for the parameters, if not stated otherwise, are the ones shown in table 1.
6.2 Experiment no. 1
Objective : To characterise the feasibility region for (b, p).
Constant parameters : n, W, C
1
, C
2
,
1
, and
2
Variable parameters : none
Control variable : b
Observed variable :
U(b), given by (42).
L(b), given by (43).
MATLAB script : TELPOL23e\scriptFigs-separades.m
Graphs : See Fig 2.
Conclusions : We see that:
For values of b greater than a threshold value b
f
min
, L(b) < U(b) holds and,
therefore, corresponding values for p can be found such that competition
results in an equilibrium.
14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
b/W
W = 95 kHz


L(b)
U(b)
Figure 2: Feasibility values for (b, p)
6.3 Experiment no. 3
Objective : To check the eect of varying the leased amount b of spectrum. We
assumed that p is agreed so that the Shapley values result for the prots.
Constant parameters : n, W, C
1
, C
2
,
1
, and
2
.
Variable parameters : none
Control variable : b
Observed variables :

U
1
and U
2

1
and
2
MATLAB script :TELPOL23e\scriptFigs-separades.m
Graphs : See Fig. 3, Fig. 4, and Fig. 5.
For the values b such that multiple equilibria result, i.e., b / [b
u
min
, b
u
max
], no value
is represented.
For the values of b such that (b, p) does not yield competitive equilibrium, i.e.,
b < b
f
min
, results for the MO correspond to the monopoly scenario.
Conclusions We see that:
With respect to the number of subscribers:
15
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.2
0.4
0.6
0.8
1
1.2
b/W

Figure 3: Fraction of MO users over the total number of subscribers ()

When the MO remains as the monopolistic operator, = 1 holds. When
the FO operator enters the market, i.e. b b
u
min
and b b
f
min
, the MO
operator looses market share as b increaseswhile b < b
u
max
.
With respect to utilities U
1
and U
2
:
As stated in section 4.3, the range of values where a unique equilibrium
results corresponds to those values with positive values of U
1
= U
2
.
The entry of FO is benecial for the users, as the utility increases as b in-
creases up to a value b
UW
= 0.5W. From this value, the utility decreases
down to zero. We may conclude from this behavior that a symmetric
equilibrium, where the FO and the MO have the same amount of re-
sources (the spectrum W/2), precludes the MO and the FO to exercise
any kind of market power. On the contrary, the level of competition is
maximum and the users derive the maximum utility from the service.
With respect to prots
1
and
2
:
Both operators increase their respective prots when the FO operator
enters the market. Furthermore, they keep increasing as the FO gets
more resources (the spectrum b) for providing service to its users. The
FOs higher eciency and the increasing payment to the MO explains
this behavior.
With respect to the social welfare SW:
A maximum is reached for a value b
SW
, which is greater than b
UW
.
Bearing in mind that the social welfare adds up the users utility and the
operators prots, this maximum is a trade o between the maximum
utility reached at b
UW
and the increasing prots.
16
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.2
0.4
0.6
0.8
1
1.2
b/W


U
1
U
2
Figure 4: Utility for subscribers for MO and FO
17
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
x 10
4
b/W

2
SW
Figure 5: Prots for MO and FO
18
The above conclusions are valid regardless the criteria used for xing p. In this
experiment, the value p that is agreed provides Shapley values for the prots. For
any other value of p, specically for p = U(b) i.e. the incumbent has full bar-
gaining powerthe prots
1
and
2
would obviously be dierent. Nevertheless,
the producer welfare and the equilibrium prices remain the same as those just
represented and discussed, and consequently the user utility, the user welfare and
the social welfare.
6.4 Experiment no. 4
Objective : To evaluate the optimum values b of leased spectrum from the point of
view of the welfare.
Constant parameters : n, C
1
, C
2
,
1
and
2
Variable parameters : none
Control variable : W
Observed variables :
Maximum value of b/W in the feasibility curve which results in a unique
competitive equilibrium
4
(b
max
/W).
Minimum value of b/W in the feasibility curve which results in a unique
competitive equilibrium (b
min
/W).
Value of b/W between b
min
/W and b
max
/W such that user welfare is max-
imised (b
UW
/W).
Value of b/W between b
min
/W and b
max
/W such that social welfare is max-
imised (b
SW
/W)
MATLAB script :TELPOL23e\script_caracteritzacio.m
Graphs : See Fig. 6
Conclusions :
We see that:
The value b
max
/W tends to the value 1, which is the case where the whole
spectrum W is leased to the FO and the prots are maximised. The con-
straint is given by the uniqueness of the equilibrium.
As regards b
min
/W, there is a lower range of values of W where the unique-
ness criteria constraints the possibility of a competitive equilibrium, b
min
=
b
u
min
> b
f
min
, and a higher range of values of W where the bargaining con-
straints the possibility, b
min
= b
f
min
> b
u
min
.
As explained in section 6.3, b
UW
is lower than b
SW
, and they almost keep
constant until b
min
increases and precludes any interior maximum b
UW
and
b
SW
to occur.
4
The feasibility region, when p is computed from (55), becomes a curve
19
8 9 10 11 12 13 14 15 16 17 18
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
W/n
b
/
W


b
max
/W
b
SW
/W
b
UW
/W
b
min
/W
Figure 6: Leased spectrum values which yield maximum user welfare and maximum
social welfare
The above results mean that the degenerate case b/W = 1, which would be
the optimum from the point of view of the producer welfare, is not always the
optimum from the point of view of either user welfare or social welfare. We
would argue then that a regulatory authority would have strong arguments
i.e., welfare enhancement to intervene by xing a maximum value b/W < 1 of
leased spectrum. And these arguments are independent on the procedure that
implements the bargaining on p.
7 Conclusions
Bearing in mind the analysis of the results conducted in the previous section, we can
conclude that:
1. Every actor, that is, users and the two operators, are better o when the FO oper-
ator enters the market which requires to be within the feasibility and uniqueness
region.
2. The regulator intervention is deemed necessary in order to restrain the incumbent
operator from leasing the whole amount of the spectrum to the entrant operator,
which will harm the users.
References
Duan, L. and J. Huang (2011). Economic viability of femtocell service provision. In
International Conference on Game Theory for Networks (GameNets), Shanghai,
China.
20
Duan, L., B. Shou, and J. Huang (2011). Capacity allocation and pricing strategies
for wireless femtocell services.
Lin, P., J. Zhang, Y. Chen, and Q. Zhang (2011). Macro-femto heterogeneous net-
work deployment and management: from business models to technical solutions.
Wireless Communications, IEEE 18(3), 6470.
Mukherjee, A. and E. Pennings (2006). Taris, licensing and market structure. Eu-
ropean Economic Review 50(7), 16991707.
Niyato, D. and E. Hossain (2008). Competitive pricing for spectrum sharing in cog-
nitive radio networks: dynamic game, ineciency of Nash equilibrium, and col-
lusion. IEEE Journal on Selected Areas in Communications 26(1), 192202.
Yi, Y., J. Zhang, Q. Zhang, and T. Jiang (2012). Spectrum leasing to femto service
provider with hybrid access.
21